How OMR processing works in CheckOMR

Upload a batch PDF from your scanner, choose an answer key, and let the pipeline run. Results appear in real time with CSV, Excel, and marked PDF exports.

That is the whole workflow — but a lot happens between upload and result. This walkthrough shows each stage of the CheckOMR (By Navodaya Study) pipeline, so you know exactly what the system does with your sheets and why the results come back so reliably.

The pipeline at a glance

1. Upload — scan the filled OMR sheets and upload them as a PDF batch.
2. Align — every page is located and aligned using ArUco markers printed on the sheet.
3. Detect — filled response bubbles are read with 99% accuracy.
4. OCR — handwritten details (name, father's name, roll number) are read by AI OCR with 99% accuracy.
5. Match — responses are compared against the answer key of the sheet's booklet code (A, B, C, D).
6. Export — results go out as Excel/CSV plus marked PDFs of every checked sheet.

Step 1 — Upload your batch PDF

Start from the dashboard and create a new batch. You can drag in a single multi-page PDF or upload several PDFs together — hundreds of sheets in one go. Live status tracking shows the batch moving from upload to processing to completion, so you always know where your results are.

If you scan with a mobile app like Adobe Scan, upload the PDF it produces — no ADF scanner machine or computer is required. Cloud-based processing means the same result accuracy whether you work from a PC or a phone.

Step 2 — Alignment with ArUco markers

Scans are never perfectly straight: pages tilt, shift, or get cropped slightly at the edges. Every CheckOMR sheet carries small printed fiducial patterns called ArUco markers. The pipeline reads these markers on every page and uses them to work out the sheet's exact position, scale and rotation.

This is why a photo from a phone and a sheet from an office ADF scanner end up being read equally well — the system measures where things actually are on your scan instead of assuming a perfect page.

Step 3 — Bubble detection

With the sheet aligned, the pipeline examines each response bubble and classifies it as filled or empty. Marked bubbles — whether dark pen or pencil — are recognised with 99% accuracy. The position of every bubble is known from the layout, so the system also knows exactly which question each filled bubble belongs to.

Step 4 — AI OCR on handwritten fields

Bubbles are only half the sheet. Student name, father's name and other handwritten details are read by CheckOMR's AI-based OCR, which reaches 99% accuracy even when capital and small letters are mixed or the handwriting is imperfect. The handwritten text and the roll-number bubbles are brought together into one clean student record.

Step 5 — Matching against your answer key

Evaluation compares every sheet's responses with the answer key of its booklet code. Because you can attach separate keys for Set A, B, C and D, mixed sets from the same exam hall are all evaluated together in one batch — no separate runs per set. You can build a custom key for your exam or use an admin preset (for example, JNVST).

What you get back

• Live status — processed count, failed count and average score update while the batch runs.
• Excel / CSV export — the full result sheet for your records, rank lists and parent reports.
• Marked PDF — a copy of each checked OMR sheet with colour-coded right and wrong marks, ready to share with students as proof of evaluation.

Why it comes back in minutes

Every stage runs in the cloud, in parallel, the moment your upload lands. That is how CheckOMR checks 300+ sheets in under 2 minutes with 99% accuracy — no dedicated scanner machine, no grading desk, no queue for the computer lab.

Ready to try it? Follow Upload your first batch, then set up your answer keys before your next mock test.